In (Old) Sindarin it seems [j] vanished before the vowel [i]. A similar sound change occurred in Welsh, where [ı̯] vanished before [i] or [e] (WGHC/§100vi). The vanishing before [e] does not occur in Noldorin or Sindarin, for example: S. eirien “daisy” (SD/126), N. iest “wish” (Ety/YES). There are a number of examples of [j] vanishing before [i], however:

• ᴹ√YER [> *jēr > *jīr] > N. îr “*desire” (EtyAC/YER).
• ᴹ√YEN [> *jēn > *jīn] > N. în “year” (Ety/YEN).
• ᴹ✶at-yēn-ar [> *atjīnar] > N. edinar “anniversary day” (Ety/YEN).

Since the combination [ji] did not occur in Primitive Elvish, this sound change only applied to cases where [i] arose later in (Old) Sindarin, such as when long [ē] became [ī]. There are also a number of Noldorin and Sindarin examples indicating [j] vanished before the diphthong [ui]:

• ᴹ√YUL > N. uilw “ember” (Ety/YUL).
• ᴹ√YŪ > N. ui- “twi-” (YŪ).
• ᴹ√YŪ > N. uial “twilight” (Ety/KAL, YŪ).
• ✶yuy(u)ŋal > S. uial “twilight” (PE17/169).

Whether this happened generally before i-diphthongs is unclear; there are not enough examples to be sure. David Salo noted both these changes, and attributed them to Old Sindarin (GS/§4.62, §4.75), but they could have happened later.

Conceptual Development: In Gnomish, the question of [j] sound changes is complicated by the fact that initial [j] usually became [g], as discussed by Roman Rausch in Historical Phonology of Goldogrin (HPG/§2.1). However, as pointed out by Rausch, there are a couple examples of [j] vanishing before [i]:

• ᴱ√YḶTḶ > G. ilt- “to yoke” with past tense galti vs. ᴱQ. yalta “yoke” (GL/50, QL/106).
• ᴱ√yṇt > G. intha- “to join to, add, increase” with past tense ganthi vs. ᴱQ. yanta- (GL/37, 51).

The second example had a variant form gintha-, so the evidence is ambiguous. Another interesting example is:

• ᴱ√ı̯el > G. Gilim “winter” (GL/38) vs. ᴱQ. yelin (yelim-) (QL/106).

Presumably the primitive form of this name was *yelim given the Qenya stem form. This seems to imply that the change of [j-] to [g-] occurred before the limited Gnomish i-affection, so [jelim] > [gelim] > [gilim]. However, in the Qenya Lexicon the root for this word is given as ᴱ√DYELE, indicating a primitive voiced palatal spirant [ʝ] rather than the semi-vowel [j], so perhaps this is the reason for the different phonetic developments. Similarly, the root for ᴱQ. yanta- “to enlarge, increase, add to” is given as ᴱ√DẎṆTṆ in the Qenya Lexicon (QL/106), which could explain Tolkien’s vacillation between gintha- and intha-.

See OS. [j] vanished before [i], [ui] for discussion.

HPG/§2.1